KR20040071597A - Systems and methods for constructing and using models of memorability in computing and communications applications - Google Patents

Abstract

PURPOSE: A system and a method for facilitating computer-based applications depending on memorability models are provided to develop and utilize the memorability models automatically capturing a personal ability for recognizing a specific event as an important landmark. CONSTITUTION: The system provides design/analysis of timeline visualization related to display for a query result based on an index of contents. A query(1120) is received by a query component(1130). The query component parses a query stored in a data storage(1140). A landmark component(1150) searches the result related information by receiving and/or accessing the data storage, landmark events, and the query component, and outputs result related data having landmark information. The landmark is automatically generated and/or defined by a user.

Description

Systems and methods for constructing and using memory models in computing and communications applications {SYSTEMS AND METHODS FOR CONSTRUCTING AND USING MODELS OF MEMORABILITY IN COMPUTING AND COMMUNICATIONS APPLICATIONS}

The present invention provides a system that recognizes a particular event as a time-critical landmark and facilitates computer-based applications according to one or more memory models that capture the ability of a person to benefit from using the landmark in navigator or review content. And to a method.

Due to global competition, there is an ever-increasing number of requests for quick access to appropriate information. For example, quick access to appropriate information can make a difference in terms of lost money for money earned in the stock market. The demand for media and journalists places a premium on getting the right information before the competition. Other industries, such as the high tech sector and counseling, need people in their industry to be in the best position at current events and trends on any market. Similarly, within client-based systems and intranets, quick access to appropriate information is indispensable for efficient maintenance in the work environment. Thus, there is a continuing need for systems and methods that facilitate rapid access to appropriate information.

The following presents a simplified summary of the invention in order to provide a basic understanding of some aspects of the invention. This summary is not an extensive overview of the invention. It is not intended to identify key / critical elements of the invention, nor to delineate the scope of the invention. Its main purpose is to present some concepts of the invention in a simplified form as a prelude to the more detailed description that is presented later.

The present invention provides a system and method for developing and using a memory model that automatically captures a person's ability to recognize an event as an important landmark in time. The memory model is a procedure for categorizing or assigning any measure of memory to an event that can be used by various computer-based applications to assist the user in processing, receiving, and / or communicating information. And policies. As one example, an event may include appointments and other annotations within a user's calendar, holidays, news stories over time, and photos, among other items. In one particular application example, the model is used to provide a personalized index that includes a landmark in time, where the use of that index can be used to browse a directory of files or other information, and search It can be used to review the results of the engine. The memory model may include a voting model, a heuristic model, a rule model, a statistical model, and / or a complimentary model, based on a pattern of forgetfulness rather than stored items. In addition, a user interface is provided and facilitates application of the model to assist the user in the retrieval and process of information. The invention also encompasses various applications and methods for installing data storage on its own, such as providing a browsable archive of important (and less important) data. For example, the data store may capture life histories (or other events) such as "My Family Biography" and "My Autobiographies".

In yet another aspect, the present invention provides a system and method that facilitates computer-based retrieval of information, in accordance with a memory model. This includes designing and analyzing timeline visualizations in connection with displaying results for queries based at least in part on an index of content. In connection with the present invention, visualization may relate to a search engine that provides a unified index of information that is exposed to a user (eg, including web pages, emails, documents, pictures, audio, etc.). The present invention develops a value that extends the base time view by adding public landmarks (eg, holidays, important news events) and / or personal landmarks (eg, photos, important calendar events). According to one particular aspect of the invention, the search results may be represented by an overview-plus-detailed timeline visualization. The summary view can show the distribution of search hits over time, and the detailed view can consider checking individual search results. The returned item can be annotated with an icon and a short description, if desired.

When searching through an individual's email, file, or web bookmark for a particular item, a person uses a variety of strategies. Although people don't remember every aspect of the item they're looking for (e.g. the exact file and file path), they're important events in their lives (e.g., their children's birthdays, foreign trips, 9/11 raids) Famous events, or the assassination of JFK) tend to remember. The present invention can use such form of contextual information to support searching through content. Interactive visualization in accordance with the present invention provides a timeline-based presentation of search results that can be pinned by public (eg news, holiday) and / or personal (eg appointment, photo) landmark events. Visualization-based indexing and retrieval systems in accordance with the present invention provide a quick and easy way to search and navigate information content, such as the text of items exposed to the user (eg, documents, visited web pages, and emails). And index the metadata.

To achieve the objects associated with the foregoing, certain illustrative aspects of the invention are described herein in connection with the following description and the annexed drawings. These aspects are implicit, but in some of a variety of ways the theory of the invention may be utilized and the invention is intended to cover all such aspects and their equivalents. Other advantages and novel features of the invention will become apparent from consideration of the following detailed description of the invention in conjunction with the drawings.

1 is a high-level schematic illustration of various memory models that may be used for computer-based applications in accordance with aspects of the present invention.

2-5 illustrate exemplary user interfaces in accordance with aspects of the present invention.

6 and 7 illustrate exemplary influence models in accordance with aspects of the present invention.

8 and 9 illustrate exemplary decision trees in accordance with aspects of the present invention.

10 illustrates exemplary display control in accordance with aspects of the present invention.

11 is a high-level schematic illustration of an exemplary system according to the present invention.

12 is a flow chart for one particular methodology in accordance with the present invention.

13 is a typical screenshot display of a timeline visualization in accordance with the present invention.

14 is a representative visualization displaying only dates to the left of the backbone of the timeline.

FIG. 15 is a representative visualization that displays landmarks (eg, holidays, news headlines, calendar appointments, and photographs of individuals) in addition to the base date.

FIG. 16 shows that the median search time with a landmark event displayed in the timeline is much faster than the median search time when using only dates to annotate the timeline in accordance with the present invention.

17 is an exemplary operating environment in accordance with the present invention.

<Explanation of symbols for the main parts of the drawings>

100: system

110: memory model

120: application

130: Voting Model

140: heuristic model

150: rule model

160: statistical model

170: Complementary Model

The invention is now described with reference to the drawings, wherein like reference numerals are used to refer to like elements throughout. In the following description, for purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the present invention. However, it will be apparent that the invention may be practiced without these details. In other instances, well-known structures and devices are shown in block diagram form in order to facilitate describing the present invention.

As used in such applications, the terms "component", "system", "model", "application" and the like refer to a computer-related entity, that is, hardware, a combination of hardware and software, software, or running software. It is to call. For example, a component may be, but is not limited to being, a process running on a processor, a processor, an object, an executable, a thread of execution, a program, and / or a computer. For illustration purposes, both an application running on a server and the server can be a component. One or more components can reside within a thread of process and / or execution, and one component can be located on one computer and / or distributed between two or more computers.

As used herein, “inference” infers or infers about the state of the system, environment, and / or user from a set of observations captured through events and / or data. The process is generally referred to. Inference can be used to identify a specific context or action, or can generate a probability distribution over states, for example. Inference can be probabilistic—ie, calculating the probability distribution over the state of interest based on consideration of data and events. Inference can also refer to techniques used to construct high-level events from events and / or data sets. Such inference results from the construction of a news event or action from a set of observed events and / or stored event data, whether the event correlates with a closed temporal approach, and whether the event and data come from a single event and data source, or You can see if it comes from multiple events and data sources.

First, referring to FIG. 1, system 100 illustrates one or more memory models that may be used for computer-based applications in accordance with aspects of the present invention. One or more memory models 110 are provided to run one or more applications 120 to assist the user in managing, retrieving, processing and / or communicating information. Memory model 110 determines various aspects of human or user memory (eg, public and / or personal memories) for one or more events 114, and the model may be based on oblivion rather than the ability to recall. Can be. It will be appreciated that memory and forgetfulness models can be used simultaneously in accordance with the present invention. In one aspect, memory model 110 uses shared voting model 130 to determine memorable items. For example, this may include asking or automatically polling a set of users to score memory for public events. In one embodiment, the memory includes at least one (or other scoring system) of salience of news stories acquired from the entire news story and of assigning a value of 1-10 to one set of users. A scalar measure for can be collected and a method of capturing memorable news stories can be obtained by averaging the scores (or other statistical processes).

One or more heuristic models 140 may be provided as memory models 110. For example, such a model 140 may use various attributes of a message and generate an informal policy that assigns scores of memory levels or deterministic categories based on a function of the attributes. As one example, the discovery function may be configured to analyze the increasing duration of the event on the calendar (or other information source) as positively affecting the memory of the event. This involves a review of the heuristic method, which involves that images or subsets of images from an image set serve as the most memorable of the image sets snapped at the event, which is the object within the screen. Synthesis, color histogram, perceived face (e.g., by automated face recognition software), and features that involve continuous and temporal relationships between pictures (e.g., the first of a set of pictures snapped to capture events) , Or near the first), a picture related to short inter-picture intervals, a photographer's excitement of the phase of the event 114, and reviewed or displayed (relatively on the picture). Images with long dwell times, edited (cropped and renamed) pictures, etc. Based on such attributes, such as attributes representing activity. Other properties for the images include automatic analysis of the image quality including, for example, focus and orientation.

At 150, one or more rule models or rules may be provided to determine the event 114. This may include rules for automatically assigning measures of memory to news stories, such as the number of news stories, persistence in the media, the number of casualties, dollar losses associated with news stories, and surprising or unusual dimensions. And attributes such as proximity and proximity to the event user (eg, the same / different country, state, city, etc.). At 160, various statistical models may be provided to model the event 114. Statistical model 160 can be used for a variety of items that focus on the use of machine learning methods that can provide a model for predicting the memory of an item, which is calendar events, holidays, news stories , And an image, based on a set of features, and the like. Statistical models 160 and processes include the use of Bayesian learning, which may generate Bayesian dependent models such as, for example, Bayesian networks, primitive Bayesian classifiers, and / or Support VectorMachines (SVMs). have. A trainer (not shown) may be provided, depending on the application, to obtain specified examples of landmark items-or items that can be forgotten, or examples identified through implicit training can be provided.

In addition, the model of memory 110 may be formulated in a complimentary manner at 170, yielding a forgetful model, and thus may be affected in application 120. Thus, the complimentary model 170 describes the use of a variant of the memory 110 model that focuses on inferring the possibility that a user will not be reminded of upcoming significant events or other related information. This model 170 may include an application 120 such as a calendar to selectively highlight information that the user is likely to forget, or to change the timing or alert of the information depending on the likelihood that the information may not be stored. ) Can use reasoning. Such models of memory and oblivion can be combined with messaging and reminding systems, for example, where context-sensitive costs and benefits are information they may need because they will not remember it (eg, peripheral devices). Or information sent to the display may be considered in an informal cost-benefit analysis or a formal decision analysis that takes into account expectations of whether, when, and how to advance steps by the reminder). To alert the user. As will be described in more detail below, the process of helping the user to view an event over time is provided for browsing the information store in the context of a set of events that is important for clearing the task of identifying a document generated over time. Can be.

Memory model 110 supports various systems, processes, and applications 120. This is a memory information-management label that labels an event or item with a numerical label or a categorical label, according to any measure of the likelihood that the item is recalled, recognized as a landmark, or most representative of the event or time. It may include using the model of the application. Such an application may utilize a mathematical function that is recalled, recognized as a landmark, or assigns a scalar measure for the significance of an event or item as the most representative of the event or time. Statistical models of memory through machine learning methods may be provided and trained implicitly, or may be provided to a specified training system that collects information about samples of events or items that are memorable or unmemorable. This may include providing real time inference or classification of the likelihood to be recognized as a landmark, or to be an event or item as the most representative of an event or time, and, more generally, another degree or Probability distributions can be provided over aspects of the systems and processes provided by the present invention.

Other applications include the use of a memory model to automatically filter heterogeneous events and content streams, for example to selectively store events for the logging of lifetime events to limit the required storage memory. In addition, the use of the memory model may be used to create a means of browsing (eg, heterogeneous events or content browsing data at different levels of temporal precision (eg, 10 years, years, months, days, hours). Hierarchical lifetime log). Another application optionally selects a picture for display around the picture fetched from the picture library, including the use of a representative landmark or memory model. Another application includes the use of representative memory landmarks and memory models, including, but not limited to, the collection of photos in a slide show over time or in time for one or more events, under constraints on the total number of slides a user wants to view. Select the set of photos at the point. In another aspect, an application may include the use of representative memory landmarks and memory models, including items (eg, picture libraries, thumbnails of graphic or photographic images displayed in files, items, or folders of documents within an operating system (eg, In order to characterize or summarize the entire contents of the MS window), a set of items (e.g., images) is optionally selected. Also note that the concept of memory provides a range of targets for each of the following learning and inferences:

Memorability: The rating by which an item is reminded or perceived.

Memorable landmark: A rating in which an item is viewed as a milestone in time.

Representative landmark: A rating in which an item serves as representing items, time periods, events, sequence of events, and the like.

As mentioned above, the complement to the memory model is the model of oblivion. Thus, the present invention can similarly train a model from data, and for an item that can combine the likelihood of an item being forgotten with a cost-benefit analysis of an expected value that is forgotten or reminded to the user about the item. Inference can be performed. General decision theoretical analysis is based on the research of Principles of Mixed-Initiative Interaction by E. Horvitz, Proceedings of CHI '99, ACM SIGCHI Conferencd on Human Factors in Computing Systems, Pittsburgh, PA, May 1999. ACM Press pp 159-166. Analyze when to come forward, with the uncertainty of the need for help described.

The present invention can use such expected utility methods, and acquires the possibility of forgetting (and memory) inferred from the memory model as the center within the calculation of expected values to remind the user. Thus, the present invention provides an anticipated-utility decision as to whether and when to present a reminder to a user for any of the given item types and contexts that take into account the cost of interruption (eg, the current cost of interruption). Can be performed. Such a model can be used for the control of alerts on reminders from the desktop and, in addition, for mobile devices, through confusion of confusion and transmission costs.

Such models can be used not only for healthy people, but also to help patients with various intellectual deficits leading to memory errors. For example, memory models built from data training may be used to predict the likelihood that a patient with Alzheimer's disease will be at a particular stage of the disease. Such a model can be combined with cost-benefit analysis as described above, and can be combined with suitable hardware to provide audiovisual cues to the user, providing an ideal reminder.

2 through 17 illustrate some exemplary interfaces using memory models in accordance with the present invention. Note that each interface shown may be provided in a variety of different settings and contexts. By way of example, the applications and / or memory models described above may be associated with desktop development tools, mail applications, calendar applications, and / or web browsers, and other types of applications may be used. These applications may be associated with a graphical user interface (GUI), where the GUI has various configurable dimensions, shapes, colors, text, data, and voice to facilitate operation with the application and / or memory model. Provides a display having one or more display objects (not shown) including aspects such as configurable icons, buttons, sliders, input boxes, selection options, menus, tabs, and the like. In addition, the GUI may also include a plurality of other inputs or controls for adjusting and configuring one or more aspects of the present invention, which will be described in detail below. It receives user commands from a mouse, keyboard, voice input, website, remote web service, pattern recognizer, face recognizer, and / or other device such as a camera or video input to affect or change the operation of the GUI. It includes a step.

2 illustrates an exemplary interface 200 employing a memory model in accordance with the present invention. Interface 200 (e.g., a memory lens) posts an event backbone on any directory being examined. Important personal events are filtered out of all available events and posted in the left column 210. Files or other data created or modified at different time points are displayed at the appropriate time period on the right column at 220. Slider 230 moves to the "most memorable" landmark, causing the landmark event from the user's calendar to be displayed, which, for the setting of slider 230, has a higher probability than the memory threshold. However, the interface 200 describes the use of an appointment item, as can be seen that similar methods can be applied to adding key images, news stories, etc., to the left column 210. The file can be launched directly from these columns (eg, a mouse click) as in other file browsers. 4 shows how the slider 300 is moved to the right (arrow direction) to add a low likelihood event that is being memory landmarked. As such, more events are added from those shown in FIG. 3. Proceeding to FIG. 4, the slider 400 moves further to the right, allowing more events to be added, i.e., the lower likelihood events under memory landmark are now included. As the slider moves, other events are added, including, for example, Ground Hog day, regular meetings with the association, and the brother's birthday. A display affordance is provided that gradually brightens events where the likelihood of becoming a landmark is progressively lower, in which case the intensity as a function of membership of the event within different ranges of the likelihood of being a landmark is provided. Allocating may be introduced.

Training systems and methods may be invoked at the interfaces described above. 5 illustrates an interface 500, where a trainer retrieves a file of a user's calendar appointments over the years and allows the user to indicate whether the appointment serves as a memory landmark. The user assigns these labels to a subset of some appointments. When the user is finished, a statistical classifier is created when the user presses the "train" button 510, which acquires several characteristics of the user's calendar events and predicts the likelihood that each event is a landmark event.

That is, p (memory landmark | E1 ... En), where p is the probability and E1 ... En is evidence related to one or more event characteristics (e.g., proximity of events to holidays, important or urgent meetings) Such as keywords, award presentation or receiving indicators, milestone meetings, performance reviews, etc.). This probability may be assigned to an unscored calendar event for use in the aforementioned interface.

Note that more than one decision model can be formulated to compute a memory model. For example, consider the model 600 shown in FIG. 6 that appears as an influence diagram. Impact diagrams are known representations of decision problems in the decision science community. This model captures uncertain relationships between key variables, including observational variables, determinations, and value functions. The impact diagram shown in FIG. 6 captures components that affect memory levels from the user's appointment, and other variable sources may also be used. In model 600, key variables (which may include other variables as well), including observational and speculative variables, are presented as elliptical nodes in graph 600. The indicated arcs indicate the probable or deterministic dependence of the variables. Model 600 represents a Bayesian network (possibility dependent model) deduced from the data. Note that the variable under consideration can be automatically collected from the user's online appointment. Some of the more interesting variables are whether or not you are (organically) insightful in the meeting, date, time, duration of the meeting, whether the meeting is regular, setting a time for advanced reminders for the meeting, and the role of the user (host? , Etc.), whether the meeting was conducted by nickname or from a person, how many people attended the meeting, whether it was reported directly by the user, whether the manager or manager's manager came to the meeting, who was the host of the meeting, the subject of the meeting, the meeting Where, how the user responded to the meeting request. Some variables under consideration in statistical modeling (see Bayesian Network Model) are specifically designed for this kind of memory landmark application. These include "Organizer atypia", "Place Ataipia", and "Attendee Attaipia". They are computed from the user's appointment store and capture the "atypia" or rarity of the attributes of the event or appointment.

Organizer Otaipia refers to the frequency with which organizers organize meetings. All appointments will be investigated and the organizer will be announced. A fraction of the time the current organizer was the organizer for the meeting is computed for each meeting being analyzed. The same is done for the place and attendees at the meeting. For the latter, the least typical attendee is considered to be the less typical attendee attribute for the event. In one embodiment, the present invention typically discretizes to a location, organizer, and attendee based on a range of frequencies, examples of which are as follows.

0% to 1%-not very typical

1% to 5%-not typical

5% to 10%-typical

10% to 100%-very typical

FIG. 7 is a diagram of some of the more important variables from a particular test set for direct dependencies, with variables indicating the likelihood that the meeting is a landmark meeting at 710. 8 illustrates a decision tree generated by a statistical modeling tool. This tree operates within the "landmark meeting" variable 710 in FIG. 9 illustrates a zoom in on the middle portion of the decision tree in FIG. 8 for predicting a landmark meeting. The length of the bar at the leaves of each set of branches or "paths" indicates the likelihood that the meeting will be considered a landmark meeting. The main branch displayed here represents a meeting that he has responded to, not being regular, not in a building, and marked as busy. Additional properties are considered in downstream branching.

10 is a display that may be selected by a user to control how / when an event and an item are displayed (eg, always, when having an event or item, when having an event, when having an item). It is a figure which shows control. The above-described interface poses some interesting design inquiries about the method and control, for the presence of documents or other items, and / or symbols for the display of apparent dates and times based on the events that were the aforementioned thresholds. In order to reformat it as more events, it is above the threshold as the slider moves, thus controlling the threshold to allow the appointment as the event backbone.

11 is a diagram of a system 1100 in accordance with one of the specific aspects of the present invention that facilitates computer-based information retrieval. System 1100 provides for the design and analysis of timeline visualizations associated with displaying results for a query based at least in part on an index of content. The query 1120 is part of a search engine that provides a query component 1130 (which includes an index of unified information exposed to the user (eg, including web pages, emails, documents, pictures, audio ...). Can be received). The query component 1130 parses the query into portions related to affecting meaningful searches in accordance with the present invention. The query component can access and locate the data store 1140, which may include the information retrieved. It will be appreciated that the data store represents a place for storing data. As such, the data storage unit 1140 may represent a distributed storage system, a plurality of heterogeneous data storage units, a single memory location, and the like. Landmark component 1150 may include data storage, and anchor public (eg, news, holidays) and / or personal (eg, appointments, photographs) landmark events to retrieve results related information. Receive and / or access information from query component 1130. Landmark component 1150 outputs result related data with landmark information at 1160. Landmarks may be automatically created and / or defined by a user. The system 1100 may index the text and metadata of items (eg, documents, visited web pages, and emails) that are exposed to the user to provide a quick and easy way to search for content. Thus, the system 1100 may display a basic time view by adding public landmarks (eg, holidays, important news events) and / or personal landmarks (eg, photos, important calendar events). ), Where the search results can be displayed as an overview-plus-detail timeline visualization.

12 illustrates a high level method 1200 in accordance with one of the specific aspects of the invention. At 1210, an inquiry is received. At 1220, the query related result data is anchored / commented with the landmark related data. At 1230, a timeline visualization is provided that displays the results for the query based at least in part on the index of the content.

Psychological literature involves a rich debate of episodic memory, which remembers the past, which may be organized by episodes, such as the location of the event, who existed, and what happened before and after the event. Contains information. Research also suggests that people use routine or exceptional events as "anchors" when trying to reconstruct past memory. The time of a particular event can be recalled by framing it in terms of other events, historical or autobiographical. Visualizations associated with the present invention harness these ideas by annotating Bayesian timelines with personal and / or public landmarks when displaying the results of a user's search for personal content.

A study of memory for computing events has shown that people have forgotten the number of many computing tasks they have performed in the past month. Their knowledge of the temporal order of these tasks also declined after a month, but when awakened by the video and photographs of their work during the target time period, they were able to remarkably remember more of the tasks they had done, I could remember the actual order more accurately. More generally, the study of specificity encoding emphasizes the interdependence between the encoded and what cues are successful later for retrieval. Memory also depends on the recovery of more general learning contexts as well as item specific contexts.

The larger body of research on efficient searching involves working with visualization search results in matrices in which rows and columns can be ordered by a number of user-specified parameters, and text and 2D interfaces are typically 3D interfaces for maximum search tasks. There is work to suggest something more efficient, and research on displaying category, summary, and / or thumbnail information with search results. The present invention uses timelines and temporary landmarks to guide the search for content (eg, personal content). Time is a common organizational structure for applications and data. Lifeline by Plaisant and others (see Lifeline by Mr. Pleasant, Mila B, Rose A, Widorf S and Schneidermann B: Visualizing Personal Histories, Proceedings of CHI 1996, 221-228) By utilizing the time-based structure of human memory by displaying the history of the individual. Kumar et al.'S Metadata Visualizing for Digital Libraries: Interactive Timeline Editing and Review, Proceedings of the 3rd ACM Conference on Digital Libraries (1998), 126-133, refer to Kumar V, Furuta Al, and Allen R. Use timelines to visualize topics such as world history and stock prices, as well as metadata about documents in libraries such as public dates. Rekimoto's "Time-Machine Computing" (Rekimoto J's Time-Machine Computing: A Time-centric Approach for the Information Environment, Proceedings of UIST 1999. "Time travel" in the version of allows users to find old documents, reversing the fact that human activities are closely related to time. LifeStream (see Rekimoto J.'s Time-Machine Computing: A Time-centric Approach for the Information Environment, Proceedings of UIST 1999, 45-54) represents a user's personal file system in a timeline format. "Forget-Me-Not" is an everywhere computing system that acts as a memory boost by gathering information about daily events from other devices in the environment and allowing the reading and filtering of these reports. Meetings with coworkers (time, place, and the name of the person attending), telephone, and email are examples of the types of data collected and are available as memory cues. "Save Everything" (see Hull J. and Hart P.'s Toward Zero Effort Personal Document Management, IEEE Computer (March 2001), 30-35) collects various pieces of data about the document and then obtains it (for example, , Fax, e-mail, copy) or similar metadata that allows for inquiries using personal metadata such as related activities that occur during the acquisition of data. Mineman and Harrison's Timestream (Space, Timestreams, and Architecture: Design in the Age of Digital Video, Proceedings of the Third International International Federation of Information Processing WG5.2 Workshop on Formal Design Methods for CAD) (1997) index into audio and video streams using all activities (e.g., speaking, sketch drawing, note typing). In contrast to these efforts, the system 1100 according to the present invention uses several personal and public landmarks as memory signals to investigate whether this context provides a useful memory prompt for efficiently retrieving a person's content. . While previous research efforts have individually examined timeline-based visualization, contextual signals for searching, or other methods for increasing search efficiency, the present invention is directed to content (e.g., personal content). All three regions are connected by using a metaphor of the timeline combined as a contextual signal in the search for.

Visualize

13 is a diagram illustrating an exemplary screenshot display of a timeline visualization in accordance with the present invention. The overview area on the left shows a timeline with hash marks indicating the distribution of search results over time. The highlighted area of the overview timeline corresponds to the segment of time displayed in the detail view. To the left of the detailed timeline backbone, basic dates as well as landmarks derived from news headlines, holidays, calendar appointments, and digital graphs provide context. To the right of the backbone, the details of the individual search results (displayed by icon and title) are presented in chronological order.

To test the value annotated on the timeline with a temporary landmark, a prototype was developed that provides an interactive visualization of the results output by the search application. The visualization shown in FIG. 13 has two main components that provide both an overview and details of the search results. The left edge of the display represents the overview timeline and their endpoints are labeled as the date of the first and last search results returned. The yearly boundary is also marked on the overview if the search results are spanned for example for more than one year. Time flows from the top of the display to the bottom, with the most recent result at the top. The overview provides the user with a general impression of the number of search results and their distribution over time. A portion of the outline is highlighted, which corresponds to the section currently focused on the detail area of the visualization. The user can interact with the overview timeline as if it were a scroll bar by selecting the highlighted area (eg with a mouse cursor) and moving it to another area of the timeline to change the portion of time displayed in the detail view. The detail portion of the visualization shows a zoom in section of the timeline corresponding to the slice of the highlighted time in the overview area. Each search result appears at the time the document was most recently stored. Not only the title of the document (or the subject line and creator in the case of an email) but also an icon indicating the type of document (html, email, word processor, etc.) is displayed. By hovering the cursor on a particular search result, the user can get more detailed information about the object, including the full path, a preview (or other amount) of the first 512 characters of the document, and the recipient, sender, and reference information in the case of a mail message. You will see a pop-up summary containing the information. Open the target item with the appropriate application by clicking on the result. The search results are displayed to the right of the backbone of the detailed timeline. The left side of the backbone is used to represent date and landmark information. The date appears closest to the backbone. The granularity (time, day of the week, month, or year) of the date shown depends on the current level of zoom. Four types of landmarks, such as holidays, news headlines, calendar appointments, and digital photos (which may include more or fewer types) may be displayed to the left of the date. Each landmark appears in a different color (which can be similar). The scale, ordering, and placement of the foregoing aspects can be tailored as appropriate for each need.

Public landmark

Public landmarks are derived from events in which the user's broad base is typically known. The landmark is given a priority class and typically only landmarks that meet the threshold priority are displayed. For the prototype according to the present invention, all users see the same public landmark, and the ratings of news headlines that different aspects of the present invention do not deem as important or memorable, for example, are important to them. By lowering it, users can be checked to customize these public landmarks.

holiday

A list of nonreligious holidays that are commonly celebrated in the United States has been obtained, and the dates of the holidays are generated from 1994 to 2004 by extracting that information from the calendar. Priority was manually assigned to each holiday based on knowledge of American culture (eg, Groundhog Day was given a low priority, while Thanksgiving was given a high priority). Holidays and priorities can be easily applied for any culture.

News headlines

News headlines from 1994-2001 were extracted from world history timelines coming into commercially available multimedia encyclopedia programs. Since the 2002 event was not available, the inventor of the present invention used the recollection of his current event to supply most of the news headlines from that year. Ten employees from the organization (none of which attended subsequent user studies) rated a set of news headlines on a one to ten scale based on their findings of how memorable these events were. The average of these scores was used to assign news landmarks to priorities.

Personal landmark

An individual landmark is unique for each user. For the prototype, all of these landmarks are automatically generated, but for other aspects of the invention it should be understood that the user may have the option of specifying his landmark.

Calendar appointments

The date, time, and title of the appointment stored in the user's calendar were automatically extracted for use as a landmark event. Promises were assigned priorities according to a set of heuristics. If the appointment is being repeated, the priority is lowered because it seems memorable and less noticeable. Appointment priorities increased in proportion to the duration of the event, and longer events (such as meetings or vacations) seem especially memorable. For the same reason, appointments displayed as "out of office" time received an increase in score. "Temporary" appointments are flagged with low priority while "important" appointments are explicitly tagged as increased priority.

Digital photograph

The prototype crawled the user's digital photograph (if he had it). The first picture taken on a given day was selected as a landmark for that day and a thumbnail (64 pixels along the longer side) was created. At the beginning of a given year, the photograph was given a higher priority than the beginning of the month, which was then ranked higher than the beginning of the day. Accordingly, an appropriate number of photo landmarks can be shown as the zoom level is changed. The inventors did not investigate more complex algorithms for selecting photos for display, but such techniques (Graham A, Gacia Morina H, Packey May and Winograd Tea's Time as Essence for Photo Browsing Through Personal Digital Libraries, Proceedings of the See Second ACM / IEEE-CS Joint Conference on Digital Libraries (2002), 326-335, or Flat Jay Auto Album's Clustering Digital Photographs Using Probabilistic Model Merging, IEEE Workshop on Content-Based Access of Image and Video Libraries 2000, 96-100. Will be considered in accordance with the invention and intended to fall within the scope of the claims appended hereto.

study

A user study was conducted to evaluate the concept behind the prototype. The goal is to learn whether timeline based presentation of search results has benefited the user and whether different types of landmarks have improved the usability of the timeline view for search. Both quantitative and qualitative data were collected to investigate these issues.

Participant

The participants were twelve employees from a group, all of whom were men from 25 to 60 years old. The prerequisite for participating in this study is the user of the search system (eg, Stuff I've Seen (SIS)).

Ready

On the day before each participant came to the usability lab, they were asked to perform two tasks. First, the inventors asked participants to install a program that extracted all the titles of their non-individual reservations from their calendars and then email the inventors this list of titles. This information has been adopted to generate two to eight personalized inquiries based on tamed speculations regarding their reservations (e.g., if they made a reservation called "Travel to Florida", Find the webpage you used when you purchased your ticket to Florida, or if they had a reservation called "CHI2002," we would have asked them to find the papers that were submitted to CHI2002.) .

Second, each participant receives a .pst file (for example, a repository of Microsoft Outlook ^TM email messages) so that the SIS application running on their machine can access the file before they arrive at the study site. You will have time to index the content. This file contains a collection of messages that have been delivered to numerous people in the organization (eg, announcements of conversations, holiday parties, promotions, etc.), which everyone would have received at some point. The inventors knew that everyone received this message because they were originally delivered to a large mailing list, but they do not know in advance whether individual participants have archived or deleted such mail. Thus, the inventor sent the participants a .pst file to facilitate the target items in their indexing.

Way

When the participants came to the available labs, they were asked to use Windows XP's remote desktop technology to access their office computers. While the participant toured the laboratory, the inventor installed the visualization client on their machine, according to the present invention. Participants first filled out inquiries asking for not only demographic information, but also information about their search and filing habits and how they stored the information. Next, they read the individual's tutorial and used the timeline interface to run two actual searches. They were given ample time to complete private lessons and were allowed to inquire. The experiment began after the individual's instruction was completed.

This experiment had a design according to the invention. Each participant was given a set of tasks to complete using two different interfaces. For half of the work, they see their search results presented in the context of a timeline annotated only with dates (FIG. 14), and for the other half they are in addition to the base date, calendar scheduling, news headlines, holidays, and You will see the timeline annotated by the digital picture (if they have saved it on their computer) (FIG. 15). The conditions are balanced to avoid learning effects, where half of the participants experience the landmark condition before the date unique condition and the other half experience the conditions in the reverse order. To avoid the order effect, the order of the doors is changed randomly for every participant pair.

The inventor used two kinds of queries. Inquiries common to 30 participants and unique personalized inquiries for every two to eight participants. The first 15 inquiries in each of the two conditions are related to finding the items that the inventors have known to a large number of employees and that they have included in the .pst file that the inventors installed the day before. For each of these 30 common tasks, the inventors provided participants with some inquiries about the issue and instructed them not to change them. The inventor chose to use a preset query because the purpose of the query was to test how well the timeline and landmarks help users navigate among their search results, and the inventor chose to use the preset query. I didn't want to inadvertently quit testing how well I could formulate. Thus, the inventors have chosen a sufficiently wide range of queries to ensure that the target item appears anywhere on the timeline, but also to produce many other results.

At the end of each set of common inquiries, the inventor asked for some inquiries that the inventor made for each user based on subset lines from their calendar reservations that the inventors extracted one day ago. While these inquiries are different for each participant, the inventors felt that they were important enough to be added because they target documents that are more personal and easier to remember than company-level email messages. For these personal tasks, the user was also allowed to enter the queries they chose and also restructure the queries to refine this search if desired.

Once the inquiry has been issued, the user can navigate through the timeline and search results by listing icons and titles, hovering over pop-up summaries for more detailed information, or clicking to open the actual document. Can be investigated. When they find the target item, they click on the large button displayed "found it" and are automatically presented with the next task and inquiry. If they cannot place the target item, there is also a button labeled "Abandon", which allows them to proceed to the next inquiry. During the experiment, the software includes all of these interactions, including information about the number of search results returned for each query, the number of different types of landmarks displayed, and the user's hovering, clicking, and overall timing of the interaction. Log the details.

After completing all the work, the participant fills out another inquiry requesting feedback on the usability of the software, the availability of the timeline presentation, and feedback on various types of landmarks, and asking for free comments.

In summary, each of the 12 study participants is presented for both of the experimental conditions, ie using a timeline with dates and landmarks, and using a date only timeline. In each condition, participants use visualizations to answer two types of inquiries: fixed inquiries about e-mail that were delivered to large distribution lists, and personalized inquiries customized for each participant.

result

Search time

Analyzes were performed on each participant's median search time to help reduce common mismatches in human behavior time. Here, the inventors looked at only common inquiries for all participants for a fair comparison. A sample test of pairs of intermediate search times for each participant indicates that the time for the landmark condition is significantly faster than the date unique condition; t (11) = 2.33, p <0.05. The comparison of the mean of the median search times is shown in FIG. 14 (standard error versus mean). For landmark conditions, the mean of the intermediate search times was 18.37 seconds, and for date unique conditions this value was 24.25 seconds. Not surprisingly, the timing data for personalized inquiries was extremely noisy, and there was no significant difference between the two conditions for these inquiries.

Inquiries

In addition to the timing data, participants completed the inquiries at the beginning and end of the experiment. Participants first entered some demographic information, followed by a number of inquiries using the 7-point Likert scale. (1 point is "strong inconsistency" and 7 points is "strong match". For example, "I like using this software" or "If you need to find old documents or e-mail, it is relatively easy to find"). Finally, the participant answered a number of free inquiries (eg, "What type of search task do you think the landmarks will help you make your searches more efficient?").

At the start of each session, before looking at the visualization, the participants answered a series of questions about their current strategy for locating the document (Table 1). The three highest ranking attributes used for search are subject, person and time. Existing search tools support access by subject and person, but lack support for time-oriented searches. This visualization, combined with a rich time display for navigation among the results, helps to heal this shortcoming by allowing keyword-based searches to generate an initial result set.

Before starting this study session, the subject is asked to rank the importance of different types of landmarks to remind them of the event. Note that public events (global events and holidays) are ranked lower than personal events. One user said, "Photos are as easy to use as calendar bookings. However, news events and holidays are less important. I mean, I know that Halloween is in October and Christmas is in December. Reminding them doesn't add information. " Another user said: "For me, an event close to my life is important, and then a global event. Of course, the 9/11 terrorist incident is a big thing for me, but what happens before I go to Africa, Think about work ".

Means of interest for future work are, in addition to the two conditions explored here, by distinguishing between different types of events dealing with "personal landmarks" and "public landmarks", date unique conditions versus all landmarks. To expand the study of conditions. After the experiment, participants evaluated the general utility of the timeline interface (Table 3). Participants, it is worthwhile to search further on which classes of search tasks are better suited for time-based presentation of results and whether other types of tasks may work optimally with alternative organizational methods. In general, we have found that time-based presentation of results is useful. One participant suggested that landmarks are most useful when "looking for time or event-related mails; finding Rick's mails for airport closures is well coupled to the September 11 terrorist attacks."

Although the vertical presentation of the timeline is well accommodated, many users have wanted the option to reverse the flow of time so that more recent search results are displayed near the bottom of the screen. This preference for time direction is often related to whether their email client displays a newer message above or below the message signal. As will be appreciated, the present invention may employ various timeline assignments (e.g., horizontal timeline, reverse timeline).

The user generally found the overview provided in the visualization useful (one user commented: "I like the way a small horizontal line shows the eruption of an activity. In this way I find out what time period it happened. Could "). However, other users thought it was confusing to navigate through the search results by selecting sections of the overview timeline (another user said, "adjusting the time scale on the overview pane doesn't seem easy for me." ).

conclusion

The inventor has developed and evaluated a timeline based visualization of search results for individual content. The results for episode memories are that they hope to add timelines to public (news headlines and Huda) and individuals (calendar) in the hope that these additions will help people locate their search goals. Reservations and digital photography) led to expansion into the Lanmark event. A user study found that searching a timeline extended with landmarks would save a significant amount of time compared to a timeline displayed only by date. In addition, the inventors have gathered significant feedback on the way users believe their events and their response to visualization, in their belief. This work demonstrates the usefulness of adding global and personal content to the presentation of search results and providing direction for future research.

At least in light of the above, the inventors consider different types of reviews when reviewing search results, more generally when timeline-focused viewing for investigation is most useful for finding results of interest. The relative value of the temporary landmark was assumed. For example, the distribution of items in time, answered for a specific query, will affect the overall usefulness of the timeline view when trying to find the item. There are also numerous other means to refine the system. The user reported some difficulties in navigating the timeline, and the inventor hoped to improve the control of the navigation through a combination of better zooming and time shifting. Thus, one particular aspect of the present invention is to create heuristics (or other models) for selecting and ranking landmarks (from all sources) and for exploring different types of summary landmarks. have. For example, painting segments of the outline timeline in different colors may be employed to display year or season within a year. The landmarks themselves pertaining to the search results can also be identified, including key attributes about the content and structure of the document, and the like. In addition to manually displaying landmarks, users can combine landmarks with more traditional search terms in the formation of inquiries, allowing users to search by "landmarks", such as "my manager and Show me all the documents I just created before reviewing the project together, or "Show me all the emails I received during the earthquake."

Referring to FIG. 17, an exemplary environment 1700 for implementing various aspects of the present invention may include a computer 1702, a computer 1702 including a processor unit 1704, a system memory 1706, and a system bus. 1708. System bus 1708 couples system elements to processing unit 1704, including but not limited to system memory 1706. Processing unit 1704 may be any of a variety of commercially available and useful processors. Dual microprocessors and other multiprocessor architectures may also be employed as the processing unit 1704.

System bus 1708 may be any of several types of bus structures, including memory buses or memory controllers, peripheral buses, and local buses, using any of a variety of commercially available bus architectures. System memory 1706 may include read only memory (ROM) 1710 and random access memory (RAM) 1712. As during startup, a basic input / output system (BIOS) is stored in the ROM 1710 that includes basic routines that assist in transferring information between elements in the computer 1702.

Computer 1702 reads hard disk drive 1714, magnetic disk drive 1716 (eg, for reading from or writing to removable disk 1718), and (eg, CD-ROM disk 1722). And an optical disk drive 1720) for reading from or writing to other optical media. The hard disk drive 1714, the magnetic disk drive 1728, and the optical disk drive 1720 are respectively characterized by a hard disk drive interface 1724, a magnetic disk drive interface 1726, and an optical disk drive interface 1728. May be connected to the system bus 1708. The drives and their associated computer readable media provide nonvolatile data storage, data structures, computer executable instructions, and the like. For computer 1702, the drive and media can accommodate storage of broadcast programming in a suitable digital format. While the above description of computer readable media refers to hard disks, removable magnetic disks and CDs, those skilled in the art will recognize that other types of media that can be read by a computer, ie zip drives, magnetic It is understood that cassettes, flash memory cards, digital video discs, cartridges, and the like can be used in the exemplary operating systems, and that any such media can also include computer executable instructions for carrying out the methods of the present invention. will be.

Multiple program modules may be stored in a drive and RAM 1712, including operating system 1730, one or more application programs 1732, other program modules 1734, and program data 1736. It will be appreciated that the present invention may be implemented as various commercially available operating systems or combinations of operating systems.

A user may enter commands and information into the computer 1702 through a keyboard 1738 and a pointing device such as a mouse 1740. Other input devices (not shown) include a microphone, IR remote control, joystick, game pad, satellite dish, scanner, and the like. These input devices are normally connected to the processing unit 1704 via a serial port interface 1742 coupled to the system bus 1708, but other interfaces such as parallel ports, game ports, universal serial bus (USB), IR interfaces, and the like. It can also be accessed through. A monitor 1744 or other type of display device is also connected to the system bus 1708 via an interface such as the video adapter 1746. In addition to the monitor 1744, the computer typically has other peripheral output devices (not shown) such as speakers, printers.

Computer 1702 can operate in a networked environment using logical connections to one or more remote computers, such as remote computer (s) 1748. Remote computer (s) 1748 may be a workstation, server computer, router, personal computer, portable computer, microprocessor-based entertainment device, peer device, or other common network node, typically May include most or all of the elements described with respect to computer 1702, but only memory storage 1750 is illustrated for simplicity. The described logical connection includes a LAN 1702 and a WAN 1754. This networking environment is common to offices, corporate wide area computer networks, intranets, and the Internet.

When used in a LAN networking environment, the computer 1702 is connected to the local network 1702 through a network interface or adapter 1756. When used in a WAN networking environment, the computer 1702 typically includes a modem 1758 or is connected to a communication server on a LAN, or has other means for establishing communication on the WAN 1754, such as the Internet. The modem 1758, which may be internal or externally mounted, is connected to the system bus 1708 via the serial port interface 1742. In a networked environment, program modules described with respect to computer 1702, or portions thereof, may be stored in remote memory storage 1750. It is to be understood that the network connections shown are exemplary and other means for establishing a communications link between the computers may be used.

According to one aspect of the invention, the filter architecture is tailored to the degree of filtering desired by the particular user of the system in which this filtering is employed. However, this custom feature applies not only to the local user system environment, but also back to the system supplier's manufacturing process where the filtering class for a particular class of users can be selected for implementation in a system produced for sale at the factory. Can be applied. For example, if the purchaser determined that the first batch of purchased systems must not be provided for the user, which should not require access to any junk mail, the factory default for this system deployment. The setting can be set high, while the second system deployment for a second class of users can be configured with a lower setting for more severe junk mail for review. In either scenario, the custom attributes of the present invention are locally enabled to allow individual users of any user class to adapt to the filtering class or, if disabled, to prevent changes to the default settings. In addition, network administrators who exercise comparable access rights to configure one or many systems suitably configured with the disclosed filter architecture can also implement this class configuration locally.

What has been described above is an example of the present invention. Of course, in order to illustrate the invention, it is not possible to describe all conceivable combinations of elements or methodologies. However, those skilled in the art will appreciate that many further combinations and combinations of the invention are possible. Accordingly, the present invention is intended to embrace all such alterations, partial alterations, and variant examples that fall within the spirit and scope of the appended claims. Furthermore, for the meaning of this term when the term 'include' is used in the description or claims, this term includes when the term 'comprising' is adopted as a transition term in the claims. It is stated here that it is used in such a way that it can include other things similar to the way 'is interpreted.

Claims (66)

In a system that facilitates computer-based searching, A query component for receiving information relating to information retrieval; Landmark component using content-based landmark information to facilitate the retrieval of the information, wherein the landmark information is stored in contextual information regarding event (s) memorable to the originator of the search. Correspondence System comprising. The system of claim 1, wherein the system provides timeline visualization in connection with displaying results for the search based at least in part on an index of the individual's content. The system of claim 1, further comprising a search engine that provides a unified index of the information exposed to the user. The system of claim 3, wherein the information comprises at least one of a web page, an email, a document, a picture, and audio. The system of claim 2, wherein the results of the search are presented in an overview-plus-detail timeline visualization. 6. The system of claim 5, further providing a summary view showing the distribution of search hits over time. 6. The system of claim 5, further providing a detailed view that allows inspection of individual search results. 8. The system of claim 7, wherein the returned items are annotated with icons and / or short descriptions. The system of claim 1, wherein the landmark component extends a basic time view by adding public landmarks and / or personal landmarks. The system of claim 1, wherein the context information is used to support searching through the context. The system of claim 1, wherein the anchoring timeline based presentation of the search using public and / or private landmark events. The system of claim 1, further comprising an indexing component capable of indexing text and / or metadata of items exposed to the user for quick and easy content retrieval. Computer-readable media having the components of claim 1 stored therein. In a method for facilitating computer-based search, Receiving information regarding information retrieval; Using content-based landmark information to facilitate the retrieval of the information, wherein the landmark information corresponds to contextual information about event (s) memorable to the originator of the search; Providing a timeline visualization of search results based at least in part on an index of the subset of contextual information How to include. 15. The method of claim 14, further comprising using one or more memory models to determine the landmark information. 16. The memory model of claim 15, wherein the memory model includes a voting model, a heuristic model, a rules model, a statistical model, an inference model, and a complimentary model. And at least one of the models. The method of claim 16, wherein the complementary model is based on a pattern of forgetfulness. 15. The method of claim 14, further comprising using the landmark information in a browser interface, wherein the browser interface associates one or more events related to the landmark information with one or more items retrievable by the browser. How. In a system that facilitates computer-based searching, Means for receiving information relating to information retrieval; Means for using content-based landmark information to facilitate the retrieval of the information, wherein the landmark information corresponds to contextual information about event (s) memorable to the originator of the search; Means for providing a timeline visualization of search results based at least in part on an index of the subset of contextual information System comprising. In systems that use memory models, One or more memory models that automatically capture a person's ability to recognize events as landmarks in time; An application that uses the memory model to facilitate the processing of information according to the events System comprising. The measure of claim 20, wherein the memory model is a measure of memorable events for use by various computer-based applications to assist users in processing, receiving, and / or communicating information. A system that includes procedures and policies for assigning memorability. The system of claim 21, wherein the events may include at least one of appointments, annotations in a user's calendar, holidays, news stories over time, and images. 21. The search engine of claim 20, wherein the memory model is used to provide a personalized index that includes a landmark in time, the index being in at least one application and in the search engine for browsing directories of information. The system used in reviewing the results. 21. The method of claim 20, wherein the memory model is one of: a voting model, a heuristic model, a rules model, a statistical model, and a complimentary model based on a pattern. A system that can include at least one. 25. The system of claim 24, wherein the voting model automatically polls a set of users to score a memorable of public events. 27. The system of claim 25, wherein the score is on a scalar scale of memorable things comprising at least one of the importance of the news stories obtained from the entire news story and the assignment of a value to one set of users. Based system. 25. The system of claim 24, wherein the heuristic model creates an informal policy that uses a characteristic of the message and assigns a deterministic category or score of memorables based on a function of the characteristic. 28. The system of claim 27, further comprising a heuristic function that analyzes the increasing duration of the events on a calendar as positively affecting the memorable of the events. 29. The method of claim 28, wherein the heuristic function is applied to image subsets or images from an image set that serve as the most memorable of the image sets obtained at the event based on one or more attributes of the images. system. 30. The image of claim 29, wherein the attribute comprises a composition of objects in the screen, a color histogram, perceived faces, sequence and temporal relationships between the pictures, and a picture related to a short inter-picture space. At least one of a capture of the photographer's excitement regarding aspects of the events, and an attribute indicative of the user's activity on the image. 31. The method of claim 30, wherein the user's activity is to examine or display the image with a longer or shorter dwell time, edit the image, crop the image, and rename it ( system including renaming). 31. The system of claim 30, further comprising an automated analysis of image sequencing, including focus and orientation. The rule model of claim 24, wherein the rule model includes a number of news stories, persistence in the media, a number of casualties, a dollar loss value associated with the news story, a surprise or atypical dimension. And a rule for automatically assigning a measure of memorable to news stories that includes an attribute relating to at least one of a feature of capturing an event and a proximity to a user of the event. 34. The method of claim 33, wherein the statistical model uses a machine learning method that provides models for predicting memorable items, wherein the statistical model is a Bayesian dependent model, primitive (such as a Bayesian network). A system comprising the use of Bayesian learning capable of generating at least one of Bayesian classifiers, and Support Vector Machines (SVMs). 25. The system of claim 24, further comprising a trainer component for obtaining explicit examples of landmark items or forgotten items. 36. The system of claim 35, wherein the trainer is supplied with examples identified through implicit training. 25. The system of claim 24, wherein the complimentary model describes the use of memorable variants focused on inferring the likelihood that a user will not be reminded of an upcoming event. 38. The method of claim 37, wherein the complementary model is used to selectively highlight information that the user is likely to forget, or to timing or alert information, depending on the likelihood that the information will not be stored. A system that uses inferences in an application to change it. 38. The messaging and reminding system of claim 37, wherein the compliance model includes context-sensitive costs and the benefit of transmitting information and alerting the user about information that may be forgotten. System coupled with. 21. The method of claim 20, further comprising threshold adjustments, wherein, depending on the setting of the adjustments, landmark events having a greater probability than a threshold of being memorable are displayed from the user's calendar. The system further includes threshold adjustment. 41. The system of claim 40, further comprising a display that gradually brightens events in which the likelihood of becoming a landmark is gradually lowered. 42. The system of claim 41 further comprising assigning intensity as a function of membership of the event within different ranges of likelihood of being a landmark. 21. The training interface of claim 20, wherein the training interface enables a user to display a file of a user's calendar appointment over the years and display whether the appointment serves as a memory landmark. The system further includes. 44. The system of claim 43, wherein the training interface further comprises a training button that takes a number of attributes of the user's calendar events and creates a statistical classifier that predicts the likelihood that each event is a landmark event. 45. The method of claim 44, wherein the probability is based on the following equation: p (memory landmark | E1 ... En), where p is a probability and E1 ... En is evidence for one or more event attributes. system. 21. The method of claim 20, wherein coworkers are in a meeting, day of the week, time of day, meeting duration, whether the meeting is regular, a time set for early reminding about the meeting, the role of the user, and the meeting is aliased. ), How many attendees are in the meeting, whether the meeting has its own reporter, manager, or manager's manager, who is the organizer of the meeting, And an inference model for processing memorability variables, including at least one of a subject of, the location of the meeting, and whether the user responded to the meeting request. 49. The "Organizer Attaipia", "Place of Claim" calculated from an appointment store of a user and capturing the rarity of an attribute of an event or appointment, i.e., "Atypia." And at least one of “Attapia” and “Attendees Atiapia”. 48. The system of claim 47, further comprising discretizing typicality for venues, hosts, and attendees into states based on a range of frequencies. 21. The system of claim 20, further comprising one or more controls selected by a user to control when and how events are displayed. In the method for applying memorable information, Labeling an event or item numerically or automatically according to a measure of the likelihood that the item is recalled, recognized as a landmark, or most representative of the event or time; Applying the labeling to information management applications How to include. 51. The method of claim 50, further comprising using a mathematical function to assign a scalar measure of the significance of an event or item as said, recognized as a landmark, or as the most representative event or time. The method of claim 51, Applying a statistical model of memorable through implicitly trained machine learning methods or using a specified training system; Collect information about a sample of memorable or non-memorable events or items that provide real time reasoning or categorization regarding the likelihood of being recalled, recognized as a landmark, or most representative of the event or time. Making a step; Providing a probability distribution over different degrees of the event or item Further comprising at least one of. 51. The method of claim 50, further comprising automatically filtering a stream of heterogeneous events or content to selectively store events for a log of lifetime events. 51. The method of claim 50, further comprising hierarchically browsing logs of heterogeneous events and content or browsing data at different levels of temporal precision. 51. The method of claim 50, further comprising using representative landmarks and memorables to selectively select pictures for an ambient display of pictures drawn from a picture library. 51. The method of claim 50, wherein in situations where the number of slides a user wishes to view is limited, selectively selecting a collection of pictures in a slide show over time or at different points in time with respect to one or more events. Using representative memory landmarks and memorables. 51. The method of claim 50, further comprising using representative memory landmarks and memorables to selectively select a collection of items to summarize the contents of the entire item or to indicate its attributes. 58. The method of claim 57, wherein the items comprise at least one of an image, a photo library, a thumbnail of graphic or photo images displayed on a file or item or document folder. 51. The information management application of claim 50 wherein the information management applications include a memorable application regarding whether the item is recalled and understood, a memorable landmark about whether the item is to be viewed as a milestone in time, and a time period of the item Or at least one of the representative landmarks as to whether they represent an event or a sequence of events. In a method for determining a reminder, Automatically training the models from the data; Performing inference on items that are likely to be forgotten How to include. The method of claim 60, Inferring the possibility of an item being forgotten; Performing a cost-benefit analysis of expectations that the user is reminded about the item How to include. 61. The expected utility of claim 60, wherein, given the item type and context, taking into account a cost of an interruption, an expected utility to determine when and when to voluntarily remind the user about something that is easy for the user to forget. The method further comprises performing an expected-utility decision. 61. The method of claim 60, further comprising controlling alerting about a reminder at a desktop application or mobile device through integration of transmission cost and disruptiveness. 61. The method of claim 60, further comprising automatically assisting patients with various cognitive deficits that can cause memory aberrancies. 65. The method of claim 64, further comprising automatically predicting the likelihood that a patient with Alzheimer's disease is at a particular stage of the disease. 66. The method of claim 65, further comprising at least one of automatically providing an audiovisual signal to users and automatically providing an ideal reminder.